Shortcut MCP Server

# Shortcut MCP Improvement Plan This document outlines planned improvements to enhance the Shortcut MCP for engineering debugging workflows with LLMs. ## Goal Enable engineers to retrieve comprehensive information from Shortcut tickets into LLM sessions that have access to source code and database information. The MCP should extract all relevant context including attachments, Loom videos, and structured data to help debug issues, request follow-ups, and update tickets with findings. ## Prioritized Improvements ### High Priority - Maximum Impact 1. **File Attachment Download Support** ✅ IMPLEMENTED - Download images, logs, and text files attached to stories - Provide content directly to LLM for analysis - Critical for accessing error traces and screenshots 2. **Loom Video Transcript Extraction** ✅ IMPLEMENTED - Extract transcripts from Loom recordings - Identify account information and user interactions - Analyze video content for debugging insights - Correlate video content with codebase files - High value for understanding reproduction steps - Note: Direct video download not available, but transcript and metadata analysis work 3. **Structured Ticket Context Builder** - Compile all story information into LLM-optimized format - Include description, comments, attachments, custom fields - Create a single comprehensive context object 4. **Linked Story Traversal** - Follow blocking/blocked relationships - Traverse linked stories for full context - Essential for complex multi-ticket issues 5. **Custom Field Extraction** - Parse account IDs, feature flags, environment details - Extract customer-specific configuration - Crucial for debugging customer issues ### Medium Priority - Significant Value 6. **Story History/Activity Feed** - Retrieve complete timeline of changes - Understand issue evolution and attempted solutions - Track who made what changes when 7. **Pull Request and Commit Linking** - Connect stories to code changes - Understand attempted fixes - Identify related code areas 8. **Intelligent Comment Summarization** - Condense long threads while preserving technical details - Highlight reproduction steps and key findings - Reduce noise in lengthy discussions 9. **Batch Story Operations** - Analyze multiple related tickets together - Pattern recognition across issues - Systematic issue identification 10. **Story Template Detection** - Extract structured data from templates - Parse bug reports and feature requests - Standardize information extraction ### Lower Priority - Nice to Have 11. **Task/Subtask Hierarchy** - Navigate parent/child relationships - Understand feature breakdown - Track completion status 12. **Story Similarity Search** - Find related issues by content - Identify duplicates - Discover patterns 13. **Export Formats** - Generate markdown for documentation - Create structured JSON for different LLM contexts - Optimize format for specific use cases 14. **Caching Layer** - Improve performance for frequently accessed stories - Reduce API calls during debugging sessions - Cache attachment content ## Implementation Notes - Focus on information extraction that directly helps debugging - Prioritize features that provide unique context not available elsewhere - Ensure all data is formatted for optimal LLM consumption - Consider rate limits and API quotas when implementing batch operations